Computer vision has quickly become one of the most common examples of the power of deep learning. There is a plethora of deep learning examples of visual recognition—many of which have a very basic level of accuracy in recognizing objects and classifying. Watson also uses deep learning for its Visual Recognition service. However, IBM has honed these algorithms, combining them with different techniques to produce a service that achieves superb accuracy, which can do so for custom recognition models with less training data, and with less compute time than many of the leading services on the market.
In this chapter, we looked at what Watson's Visual Recognition service can do out the box; how to customize it with your own set of classifiers and training data; and how to program it.
In the next chapter, we will examine the speech recognition service.